mirror of
https://github.com/deepseek-ai/deepseek-harness
synced 2026-08-15 21:04:50 +00:00
fix(workspace-context): recompose the baseline on resume, skip only on remount
The mount-local baseline guard was seeded from "a baseline already exists in the log", which a resumed session and a hot plugin remount both satisfy. That made a resume skip its baseline, so offline AGENTS.md edits or removals never reached the first resumed request — violating the documented resume contract. Distinguish the two by agent/session-start: a startup or resume emits it before the first step, while a remount attaches to an already-live session and never witnesses it. Only a remount (no witnessed start, baseline already logged) keeps the single logged baseline and skips; a resume falls through and re-composes from current files. Adds a regression that resumes a session with an offline baseline edit and asserts the fresh baseline reflects it.
This commit is contained in:
@@ -2,5 +2,5 @@
|
||||
# side as of the last confirmed-consistent state. Both languages carry equal authority;
|
||||
# after editing either side, bring the other along and re-record with:
|
||||
# pnpm run verify-translation-pairing --write
|
||||
2026-06-24-workspace-context.md: c1539840b404f8601c4a98917c8e0ca31257f688
|
||||
2026-06-24-workspace-context.zh.md: 704a369c38a83c19d5a37faf54eab0c66e0e464f
|
||||
2026-06-24-workspace-context.md: 9591e1ab169f1883e1831b50308c86e55599296b
|
||||
2026-06-24-workspace-context.zh.md: 6317ad6e2e6c85768e74b5b2657903723ddf72d3
|
||||
|
||||
@@ -32,7 +32,7 @@ At the first `agent/step` of an agent-loop instance, the plugin injects one sour
|
||||
|
||||
The injection becomes a durable `user/message` with plugin provenance. In the product spine workspace instructions are registered before the skills catalog, so their `agent/step` listener injects first. The loop drains both messages before deriving the first request.
|
||||
|
||||
A resumed agent creates a new loop instance and injects a baseline composed from current files before its first request. This permits current baseline content on resume without mutating an earlier history event.
|
||||
A resumed agent creates a new loop instance and injects a baseline composed from current files before its first request. This permits current baseline content on resume without mutating an earlier history event. A resume and a hot plugin remount both face a log that already holds a baseline; they are told apart by `agent/session-start`, which a startup or resume emits before the first step while a remount attaches to an already-live session and never sees it. Only a remount (no witnessed lifecycle start, baseline already logged) keeps the existing baseline and skips; a resume re-composes.
|
||||
|
||||
The baseline is a user-role `<system-reminder>` with `Instructions from: <path>` sections and explicit authority and precedence language. This familiar model-facing frame avoids a harness-specific XML vocabulary. Project paths are root-relative and the user-global path is `~/.dsh/AGENTS.md` for the default home or `$DSH_HOME/AGENTS.md` for a configured home. A literal `</system-reminder>` inside file content is escaped. The package README owns the exact current [prompt shape](../../../../packages/context/workspace-context/README.md#prompt-shape).
|
||||
|
||||
|
||||
@@ -32,7 +32,7 @@ Status: implemented
|
||||
|
||||
该注入成为一条带插件来源的持久 `user/message`。在产品主干中,工作区指令的注册先于 skill 目录,所以其 `agent/step` 监听器先注入。循环会在派生第一次请求前 drain 这两条消息。
|
||||
|
||||
恢复 agent 会创建新的循环实例,并在其第一次请求前注入由当前文件组合的基线。这样,恢复时可以使用当前基线内容,而无需修改先前的历史事件。
|
||||
恢复 agent 会创建新的循环实例,并在其第一次请求前注入由当前文件组合的基线。这样,恢复时可以使用当前基线内容,而无需修改先前的历史事件。恢复与插件热重挂都会面对日志中已存在基线的情况;二者通过 `agent/session-start` 区分:启动或恢复会在第一步前发出该事件,而热重挂附着到一个已存活的会话、永远不会看到它。只有热重挂(未见证生命周期开始、且日志已有基线)才保留既有基线并跳过;恢复则重新组合。
|
||||
|
||||
基线是一条 user 角色的 `<system-reminder>`,包含 `Instructions from: <path>` 章节,以及明确的权威性与优先级说明。这种熟悉的模型可见框架避免引入 harness 专用的 XML 词汇。项目路径相对于根目录;使用默认 home 时,用户全局路径为 `~/.dsh/AGENTS.md`,使用已配置 home 时则为 `$DSH_HOME/AGENTS.md`。文件内容中的字面量 `</system-reminder>` 会被转义。包 README 负责规定当前准确的[提示词形态](../../../../packages/context/workspace-context/README.md#prompt-shape)。
|
||||
|
||||
|
||||
@@ -50,23 +50,35 @@ export function apply(ctx: Context, config: Config): void {
|
||||
const instructionVersions: InstructionVersionCache = new WeakMap()
|
||||
const pendingVersionUpdates = new Map<ToolExecutionToken, InstructionVersionUpdate[]>()
|
||||
const baselineLoaded = new WeakSet<object>()
|
||||
// Sessions whose lifecycle start this mount witnessed. A startup or resume
|
||||
// emits agent/session-start before the first step; a hot remount attaches to
|
||||
// an already-live session and never sees it. That difference is the only
|
||||
// reliable way to tell a resumed session (re-compose the baseline from
|
||||
// current files) from a remount over a live one (keep the single baseline
|
||||
// already in the log) — the durable log looks identical in both cases.
|
||||
const lifecycleWitnessed = new WeakSet<object>()
|
||||
const pendingByParent = new Map<ToolExecutionToken, {
|
||||
agent: Agent
|
||||
changes: WorkspaceInstructionChange[]
|
||||
versionUpdates: InstructionVersionUpdate[]
|
||||
}>()
|
||||
|
||||
ctx.on('agent/session-start', (agent: Agent) => {
|
||||
lifecycleWitnessed.add(agent.session)
|
||||
})
|
||||
|
||||
ctx.on('session/event', (session, event) => {
|
||||
observeInstructionSessionEvent(session, event, pendingNestedChanges, instructionVersions)
|
||||
})
|
||||
|
||||
ctx.on('agent/step', async (agent: Agent, _turn, _step, signal): Promise<void> => {
|
||||
if (baselineLoaded.has(agent.session)) return
|
||||
// The guard is mount-local, but the baseline is durable: a hot remount
|
||||
// over a live session must fold the already-appended baseline from the
|
||||
// log instead of injecting a duplicate.
|
||||
if (agent.session.events.some(event => event.type === 'user/message'
|
||||
&& event.data.source.kind === 'plugin' && event.data.source.plugin === 'workspace-context')) {
|
||||
// A baseline already in the log with no witnessed lifecycle start is a hot
|
||||
// remount over a live session: keep that single baseline and skip. A
|
||||
// resumed session witnessed its start, so it falls through and re-composes.
|
||||
if (!lifecycleWitnessed.has(agent.session)
|
||||
&& agent.session.events.some(event => event.type === 'user/message'
|
||||
&& event.data.source.kind === 'plugin' && event.data.source.plugin === 'workspace-context')) {
|
||||
baselineLoaded.add(agent.session)
|
||||
return
|
||||
}
|
||||
|
||||
@@ -1005,6 +1005,47 @@ describe('workspace context request injection', () => {
|
||||
}
|
||||
})
|
||||
|
||||
it('recomposes the baseline from current files when a resumed session edited it offline', async () => {
|
||||
const root = await tempRepo()
|
||||
const home = await tempRepo()
|
||||
try {
|
||||
await mkdir(join(root, '.git'), { recursive: true })
|
||||
await write(join(root, 'AGENTS.md'), 'old root rule')
|
||||
const ctx = new Context()
|
||||
await mountWorkspaceContext(ctx, { dshHome: home, maxBytes: 65536 })
|
||||
const original = stubAgent(root)
|
||||
await composeBaselinePrefix(ctx, original)
|
||||
|
||||
// Offline edit to the baseline file, then resume on a fresh session whose
|
||||
// seeded log already carries the original baseline. A resumed session is
|
||||
// registered after this mount's apply(), so the remount guard never seeds
|
||||
// it: its first step re-composes a fresh baseline from current files,
|
||||
// reflecting the offline edit before the first resumed request. The old
|
||||
// baseline stays in history unmutated (note: resume without mutating an
|
||||
// earlier history event).
|
||||
await write(join(root, 'AGENTS.md'), 'new root rule after offline edit')
|
||||
const resumed = stubAgent(root, [...original.session.events])
|
||||
|
||||
// Resume announces its lifecycle start before the first step.
|
||||
agentEvents(ctx, resumed).emit('agent/session-start', 'resume')
|
||||
await composeBaselinePrefix(ctx, resumed)
|
||||
|
||||
const pluginMessages = resumed.session.events.filter(event =>
|
||||
event.type === 'user/message' && event.data.source.kind === 'plugin'
|
||||
&& event.data.source.plugin === 'workspace-context')
|
||||
expect(pluginMessages).toHaveLength(2)
|
||||
const latest = pluginMessages.at(-1)
|
||||
expect(latest?.type === 'user/message' && blocksText(latest.data.content))
|
||||
.toContain('new root rule after offline edit')
|
||||
const original0 = pluginMessages[0]
|
||||
expect(original0?.type === 'user/message' && blocksText(original0.data.content))
|
||||
.toContain('old root rule')
|
||||
} finally {
|
||||
await rm(root, { recursive: true, force: true })
|
||||
await rm(home, { recursive: true, force: true })
|
||||
}
|
||||
})
|
||||
|
||||
it('tracks only baseline files that were actually included under the byte budget', async () => {
|
||||
const root = await tempRepo()
|
||||
const home = await tempRepo()
|
||||
|
||||
Reference in New Issue
Block a user